Vertical Video Framing: Shooting 9:16 for AI Films

Previously: Tip #3: end every episode on a cliffhanger.

A vertical frame isn't a narrow window into a horizontal film. It changes where you put people, how you show a relationship and which details can carry the story. A beautiful wide shot can become two half-faces and an empty wall when cropped. Before generating or shooting your pilot, open a 9:16 canvas and decide what the viewer actually needs to see.

Compose the conflict vertically

Two people standing far apart work naturally in widescreen. On a phone, try arranging them in depth: one close to the camera, one in the doorway behind her. The nearer face gives us an emotion; the person behind gives us the cause. You can understand the relationship without needing the whole room.

Height can help too. A seated character looking up at someone who is blocking the exit creates a clear imbalance. A hand lowers a contract into the bottom of the frame while a face remains above it. Think about foreground, background and height, not just squeezing your subjects toward the centre.

Give faces and hands a job

Faces tell us what a discovery means. Hands tell us what a character does about it. If your scene depends on a hidden ring, a stolen key or a signature, those details need their own readable shots. A tiny object at the far edge won't become important just because you mentioned it in the script.

For example, show a hand removing a wedding ring, then the person's face as someone enters. Don't make us search a busy full-body shot for the ring. Keep the eyeline and the direction of movement consistent between shots so the viewer understands who is looking at whom. A close-up is useful when it carries information, not simply because vertical video allows one.

Leave room for captions and the interface

The uploaded picture won't always appear alone. Captions may cover the lower part of the frame, and a viewing app can place buttons along the side. The exact arrangement varies. Check the platform you're submitting to instead of assuming one universal safe-zone measurement.

In your edit, add temporary caption blocks and a rough interface overlay. Watch every important action with them visible. If a caption hides the signature, move the shot or the caption. If a reaction sits behind a column of buttons, reframe it. Keep essential faces, hands and story objects comfortably inside the available picture, with breathing room rather than touching an edge.

Plan a simple vertical shot sequence

Imagine a daughter discovering that her mother signed her name. You don't need to show the office, the street and every person in the building. Try these four shots:

• A vertical medium shot: the daughter in front, her mother waiting behind the desk.
• An insert: the daughter's finger stops at the signature on the document.
• A close-up: recognition on the daughter's face, with space below for the caption.
• A reverse shot: the mother closes the door instead of answering.

Each frame contributes a new piece of the conflict. The geography is simple, the paper is legible and the final action creates a question. You can make a tense scene with very little coverage if the shots are doing specific jobs.

Generate or shoot for the final frame

If you're using AI, specify vertical composition in the source image and inspect the result before animating it. A 9:16 output can still have a widescreen-minded arrangement, with important characters pushed to opposite edges. The aspect ratio setting alone doesn't compose the scene for you.

If you're shooting, monitor the actual vertical frame. Don't assume a later centre crop will preserve the hand entering from the side or the person waiting in the background. Leave enough room for small stabilisation adjustments, but make the intended composition in camera. Rebuilding a broken shot is usually slower than checking it first.

Do the final check on a real phone

Export a rough version and watch it at ordinary phone size, without zooming. Can you recognise the face, read the important object and follow the exchange? Watch again with captions, then muted. Ask a friend what detail they noticed in the final shot. If they missed the clue, make it larger or give it a clearer moment. For PilotRush, the frame should help the story arrive, not ask the viewer to hunt for it.